Good Day,

I have a Small Office Network running on a Miktrotik Hex series Router (RB750Gr3).

I have configured 2 WAN interfaces as I have one static IP internet ISP and another connection over PPPoE.

The third Port connects to switch which has LAN users connected via physical Ethernet and Wifi AP's.

I have setup connection marking and special routing to allow both WAN connections to be used to access the internet and if one connection fails traffic will route out of the available connection.

I have +-50 users devices that might access the LAN and I need a way to ensure they share bandwidth correctly.

Can someone point me in the correct direction such that if there is no network usage then users get full bandwidth. BUT if we are running at network throughput capacity then users will only get what is available or share evenly.

I read somewhere that simple queue can be used. But I understand this limits each user to a maximum of Throughput/users.

However I want a user to get full required speed at anytime IF there is capacity and only share what needs to be shared when there is contention.

So lets say network supports 100Mbps download.

If User1 is downloading and needs 100Mbps they should get the full 100Mbps.
If User2 starts a download but only needs 20Mbps then User1 should be throttled to 80Mbps and User2 gets full 20Mbps
If User3 starts a download but only needs 30Mbps, Then User 3 should get 30Mbps, User 2 should keep his 20Mbps and User1 now restricted to 50Mbps
If User4 starts a download and needs 50Mbps then User 2 should get his 20, User 1,3,4 should share the 80/3= 26,6Mbps as we cannot give any of them what they need.

Can I achieve this on Router OS?

Thanks

More or less you can. There's the limit-at which is the guaranteed bandwidth the user has to get, and there is the max-limit which is the maximum bandwidth the user can get if enough unused bandwidth is available. The total available bandwidth is determined by the parent queue, and the priority attribute of "leaf" queues determines the order in which the queues get a chance to grab their share of the unused bandwidth.

But there are many factors that may break this - you can only throttle the download direction indirectly, by moderating the bandwidth of forwarding received packets to their local destinations, so if the remote source doesn't accept any feedback about "how it goes", it may exhaust the download bandwidth; in the upload direction, the access network may be overbooked so you actually have less bandwidth than you expect.